Article Preview
TopIntroduction
The agriculture field has become an eminent research area for real data analysis combined with machine learning and computer vision techniques. Recently, the machine learning concept is expanded everywhere, including interdisciplinary applications. Some of the useful methods used by the researchers such as Support Vector Machine, Transfer Learning, Clustering, and Visualization Techniques. Specifically, computer vision co-relation with agriculture computes very high performance for statistical production growth of various crops. Rice plant is an important food grain for regular commentary, and it can be observed through height evaluation of plant data.
The main contribution of the proposed work is: It exhibits plant phenotyping applications such as height measurement of a single and field plants by using pixel-based clustering technique. It is also useful for disease detection at an early stage of plant data by regular observation. The proposed scheme is a hardware-free technique that avoids the complexity of the calibration setup, and it gives an easy method to calculate the height with less time and error. The heterogeneous dataset is supposed to have more noise because of environmental effects, and this work resolves those problems by using a color conversion technique. It is more useful for the massive farming or field farming analysis by calculating average height, which is not yet implemented in recent works. The proposed approach provokes digitized evaluation of real datasets with less possibility of human error rate. The primary significance of this work is the utility of machine learning combined with the image processing technique. This proposed combination gives 97.58% accuracy from the previous growth evaluation results, e.g. percentage error of 17.25% for height calculation discussed by Constantino et al. (2015). Height calculation is implemented by many methods, such as skeletonization technique. The hardware detected red and green band technique proposed by Constantino et al. (2018) with ground truth data, contour-based masking technique and feature fusion-based approach (Patel & Sharaff, 2020). Furthermore, a short discussion is listed here according to plant phenotyping (Patel & Sharaff, 2019) for growth analysis of rice crop:
Rice crop life cycle is within 120 days from the plantation to the grain filling. There are many stages which have very influential factors for the growth production, such as germination seed quality analysis, Leaf emergence, Tiller observation, vegetation growth and panicle growth analysis, Plant height observation, Panicle dry weight analysis (Best approach for market growth prediction), Dark respiration, Grain fissuring analysis, and Biomass calculation (Cai et al., 2018), etc. There are many more rice production analysis methods, but most of the work is related to disease detection using image processing techniques. This method provides an automatic scale based observation called a pixel mapping technique exclusively for height calculation.